Rubbish Check
BBC Business · 9 September 2026
source“Anthropic researcher believes more than 10% chance AI ‘could kill all humans’”
R2/ 10
Lightly altered
Rubbish Rating — 1 = base fact, 10 = pure rubbish
12345678910
In short
Rubbish Talk rates BBC's headline that an Anthropic safety researcher put "more than 10% chance" on AI killing all humans a 2/10, because the figure and quote are directly and accurately attributed to Evan Hubinger's own words, correctly framed as a personal researcher opinion rather than an Anthropic corporate position.
The Verdict
Lightly altered. The headline is a faithful, attributed paraphrase of a real statement by a named individual, not a fabricated or exaggerated claim; the only iteration away from pure base fact is that it compresses a hedged, decade-out, contingent forecast into a punchier standalone line without the "within the next decade" and "if AI can self-improve" qualifiers.
What actually happened
Evan Hubinger, an AI alignment researcher at Anthropic, posted on X that he sees a greater than 10% chance advanced AI could pose a species-ending risk within the next decade, while stressing that the risk from the models which currently exist was "low" but he was "worried" the technology might become able to improve itself soon to the point where it posed an existential risk to humanity. His remarks came in reply to a former Anthropic researcher's public warning, and Anthropic's own August safety report separately flagged early signs of rising risk without endorsing a specific probability.
Key facts
- Hubinger's stated risk applies within the next decade, and he described current-model risk as "low", not the 10%+ figure the headline foregrounds.
- Hubinger's post has been viewed more than 10 million times, and in it he said "we really do earnestly believe" AI poses a species-ending risk to humans, adding "I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence".
- Anthropic's own institutional stance is more measured: its August safety report said there was a low risk of highly-capable AI performing automated research and development that could cause catastrophic harm, though it was "less confident in this assessment" than previously and was "seeing early signs of potential acceleration".
- The claim originated from an individual's social-media post, not a peer-reviewed study or company statement, a distinction the headline preserves by naming him as "a researcher."
What to watch for
Watch whether Anthropic issues an official response distancing itself from Hubinger's personal probability estimate, and whether the separate FT report on the model being withheld from the UK's AI Safety Institute (mentioned in the same article) gets its own headline treatment, since conflating the two stories in future coverage would be a genuine framing problem.
About this scoreThe R-Score is Rubbish Talk's editorial opinion on how far a headline's framing sits from what the underlying facts support. It is a judgement about presentation and emphasis, not an allegation that any outlet has acted dishonestly. Every figure we rely on is linked under Receipts so you can check it yourself.
